# perspectives [Health: Active]

**Category:** 🗣️ Conversational AI  
**Repository:** https://github.com/polydev-ai/perspectives-mcp  
**GitHub Stars:** 0  
**npm Downloads (last month):** 1175  
**Views:** 2  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/perspectives

## Description
Query multiple AI models (GPT-4, Claude, Gemini, Grok) in parallel for diverse perspectives

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "perspectives": {
    "command": "npx",
    "args": ["-y","polydev-ai@latest"],
    "env": {
      "POLYDEV_USER_TOKEN": ""
    }
  }
}
```

**Requires environment variables:** `POLYDEV_USER_TOKEN` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What perspectives does

The perspectives MCP server gives an MCP client access to Polydev’s multi-model consultation service. Its main operation, `get_perspectives`, accepts a prompt and queries four models in parallel: OpenAI GPT 5.2, Anthropic Claude Opus 4.5, Google Gemini 3 Pro, and xAI Grok 4.1. The returned analysis presents each model’s view so an agent or developer can compare reasoning, code suggestions, and recommendations.

The supported models are described with different areas of emphasis. GPT 5.2 is listed for reasoning and code generation, Claude Opus 4.5 for analysis and nuanced thinking, Gemini 3 Pro for multimodal work and large context, and Grok 4.1 for real-time knowledge and directness.

## How it works

The perspectives MCP server is launched with `npx -y polydev-ai@latest` and communicates with compatible MCP clients over the client’s configured local process connection. A request can be made through natural language by mentioning “polydev” or “perspectives,” or by calling `get_perspectives` directly with a prompt and user token.

The service performs the multi-model query through Polydev rather than requiring separate provider credentials in the MCP configuration.

## Setup and configuration

Create a Polydev API token from the Polydev MCP tokens dashboard before starting the server. The token is supplied as `POLYDEV_USER_TOKEN`, either as an exported environment variable or in the MCP client’s server configuration.

The README provides configuration examples for Claude Code, Cursor, Windsurf, Cline in VS Code, and OpenAI Codex CLI. Codex examples also set a 180-second tool timeout and a 600-second session timeout. Polydev lists a free tier with 1,000 messages per month and a Pro tier with 10,000 messages per month at $19 per month.

## Tools and capabilities

The documented tools are:

- `get_perspectives`: Query the supported AI models simultaneously.
- `get_cli_status`: Check the status of local CLI tools.
- `force_cli_detection`: Detect installed CLI tools again.
- `send_cli_prompt`: Send a prompt to local command-line interfaces with fallback.

The multi-model response can expose separate model analyses and may provide a consensus or recommendation when the responses converge. The local CLI tools extend the server beyond hosted model consultation by allowing prompts to be sent to detected local CLIs.

## Limitations and notes

Access requires a Polydev user token and is subject to the account’s message allowance.

The README reports typical response times of 10–40 seconds and presents research results for a specific SWE-bench consultation approach. Those figures are context for the project, not a guarantee for every prompt or environment. Local CLI behavior depends on which command-line tools are installed and detected on the host.

_Full upstream README: https://allmcps.com/mcp/perspectives/readme_